Breakdown processes in HID lamps : exploration of various key aspects

(1)

Breakdown processes in HID lamps : exploration of various

key aspects

Citation for published version (APA):

Sobota, A. (2011). Breakdown processes in HID lamps : exploration of various key aspects. Technische Universiteit Eindhoven. https://doi.org/10.6100/IR708985

DOI:

10.6100/IR708985

Document status and date: Published: 01/01/2011 Document Version:

Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers) Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website.

• The final author version and the galley proof are versions of the publication after peer review.

• The final published version features the final layout of the paper including the volume, issue and page numbers.

Link to publication

General rights

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain

• You may freely distribute the URL identifying the publication in the public portal.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:

www.tue.nl/taverne

Take down policy

If you believe that this document breaches copyright please contact us at:

openaccess@tue.nl

providing details and we will investigate your claim.

(2)

Breakdown processes

in HID lamps

Exploration of various key aspects

PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de Technische Universiteit Eindhoven,

op gezag van de rector magnificus, prof.dr.ir. C.J. van Duijn, voor een commissie aangewezen door het College voor Promoties

in het openbaar te verdedigen op maandag 11 april 2011 om 16.00 uur

door

Ana Sobota

(3)

Dit proefschrift is goedgekeurd door de promotoren:

prof.dr.ir. M. Haverlag en

prof.dr.ir. G.M.W. Kroesen Copromotor:

dr.ir. J. van Dijk

This work was financially supported by Philips Lighting. CIP-DATA TECHNISCHE UNIVERSITEIT EINDHOVEN Sobota, Ana

Breakdown processes in HID lamps - Exploration of various key aspects / by Ana Sobota. - Eindhoven : Technische Universiteit Eindhoven, 2011. - Proefschrift. A catalogue record is available from the Eindhoven University of Technology Library. ISBN: 978-90-386-2453-2

NUR 924

All rights reserved. No part of this book may be reproduced, stored in a database or retrieval system, or published, in any form or in any way, electronically, mechanically, by print, photo-print, microfilm or any other means without prior written permission of the author. Printed by Ipskamp Drukkers B.V.

(4)

Summary

This thesis presents the results of experimental and modelling studies of break-down processes in near-atmospheric pressure noble gasses. The motivation came from the lighting industry - our goal was to provide a better understand-ing of the breakdown phenomena in conditions typical for a mid-pressure high-intensity discharge (HID) lamp. However, many parts of the research can be used in a broader spectrum of applications involving breakdown, for example in high power electronics, in removal of unwanted electric charges, photocopying, handling waste, UV generation or surface treatment. We focused our research to mid-pressure (0.1 to 1 bar) discharges in argon and xenon with varying conditions that will prove to greatly influence the breakdown process.

First, we examined the effect the dielectric surfaces have on the breakdown process. A pin-to-pin electrode geometry was placed in close vicinity of a flat dielectric in an argon atmosphere of pressure varying between 0.1 and 1 bar. We used positive pulsed voltages on the charged electrode (rise time varying between 47 and 100 V/ns), observed the development and measured the speed of the discharge forming on the dielectric surface and in the gas between the electrode tips. Our results show that surface discharges use propagation and growth mechanisms that are in some aspects different from the discharges that form in the gas.

The effect of the voltage form on the breakdown process was subsequently studied. Lowering of the breakdown voltage of lamps is a constant goal to be met, and it has already been observed that substituting pulsed voltages for AC in the 100-kHz range brings significant improvements. We performed electrical and optical measurements of the breakdown parameters and explained why AC breakdown works on lower voltages than pulsed breakdown. The differences between the discharges in different gasses were explained, along with the influ-iii

(5)

ence of voltage frequency on the breakdown process and UV- and85_Kr-related

effects. Statistical lag times were calculated for different parameters.

A valuable contribution to the understanding of the AC ignition process was done by using computer simulations. A fluid model and a cylindrically symmetric 2D geometry with a pin-to-pin electrode configuration was used to simulate a 700 mbar argon discharge. After simulating pulsed ignition in free gas and proving the importance of metastables on the discharge growth, the AC breakdown process in a lamp-like geometry was also simulated at frequencies between 60 kHz and 1 MHz. The main finding of this part of the research was the explanation why AC breakdown requires lower voltage than pulsed breakdown. We also explained the influence of the voltage frequency observed in experiments.

The final part of the research considered the influence of external structures (“antennas”) on the breakdown process in AC discharges. Antennas are thin metallic formations on the outer surface of the lamp burner. An EM model was used to examine the influence of different antenna structures on the electric field enhancement in the lamp in a static case. We have also done a series of experiments on lamps, showing that the antennas significantly lower the breakdown voltage. The last part of the thesis shows how antennas work, why the active ones work better than the passive ones, and the reason behind the observed differences in the workings of the passive antennas.

(6)

Samenvatting

Dit proefschrift behandelt experimenteel en modelmatig onderzoek aan doorslag-processen in edelgassen onder atmosferische druk. De motivatie hiervoor kwam vanuit de verlichtingsindustrie - ons doel was om een beter begrip te krij-gen van de doorslagprocessen in omstandigheden die representatief zijn voor midden-druk High Intensity Discharge (HID) lampen. Echter, de resultaten van het onderzoek kunnen ook worden gebruikt in een breder spectrum van toepassingen, bijvoorbeeld in de hoogvermogenselektronica, in de verwijdering van ongewenste elektrische ladingen, fotokopi¨eren, behandeling van afval, UV-opwekking of oppervlaktebehandeling. We hebben ons onderzoek in 0.1 tot 1 bar ontladingen in argon en xenon gedaan, onder wisselende omstandigheden die grote invloed op het doorslagproces zullen blijken te hebben.

Eerst onderzochten we het effect van de diëlektrische oppervlakken op het doorslagproces. Een pin-to-pin elektrode geometrie werd geplaatst vlakbij een vlak diëlektricum. Het experiment werd in een argon atmosfeer gedaan, bij een druk die varieerde tussen 0.1 en 1 bar. We hebben positieve gepulste spanningen op de bekrachtigde elektrode gebruikt. De snelheid van de ont-ladingen die zich vormden op het diëlektrisch oppervlak en in het gas tussen de elektrodes werd ook gemeten. Onze resultaten laten zien dat de ontladingen op oppervlakken enkele groeimechanismen gebruiken die in sommige opzichten verschillen van de ontladingen die zich in het gas vormen.

Vervolgens werd het effect van de spanningsvorm op het doorslagproces on-derzocht. Verlaging van de ontsteekspanning van lampen is een voortdurende doelstelling, en het is reeds vastgesteld dat vervanging van gepulste spanningen door AC spanningen substanti¨ele verbeteringen met zich meebrengt. We voer-den elektrische en optische metingen uit en hebben kunnen uitleggen waarom AC ontsteking bij lagere spanningen mogelijk is dan gepulste ontsteking. De v

(7)

verschillen tussen de ontladingen in de verschillende gassen werden uitgelegd, samen met de invloed van de spanningsfrequentie op het doorslagproces en UV- en 85_{Kr-gerelateerde effecten. Stochastische vertragingseffecten werden}

berekend voor verschillende parameters.

We hebben een waardevolle bijdrage geleverd aan het begrip van het AC ontstekingproces door middel van computersimulaties. Een vloeistofmodel en een 2D-geometrie met cylindrische symmetrie met een pin-to-pin configuratie van elektroden werden gebruikt om een 700 mbar argon ontlading te simuleren. De simulaties van deze gepulste elektrische ontsteking hebben het belang van metastabiele atomen aangetoond. De AC doorslag in een lamp-achtige ge-ometrie is ook gesimuleerd bij frequenties tussen 60 kHz en 1 MHz. De be-langrijkste bevinding van dit deel van het onderzoek was de verklaring waarom AC ontsteking een lagere spanning vereist dan gepulste ontsteking. We hebben ook een beter begrip van de invloed van de spanningsfrequentie in experimenten gekregen.

Het laatste deel van het onderzoek gaat over de invloed van externe struc-turen (“antennes”) op het doorslagproces in AC ontladingen. Antennes zijn dunne metalen structuren aan de buitenkant van de ontladingsballon. Een EM-model werd gebruikt om de invloed van verschillende antennestructuren te onderzoeken. We hebben ook experimenten aan lampen gedaan, waaruit blijkt dat de antennes de doorslagspanning aanzienlijk verlagen. Het laatste deel van het proefschrift laat zien hoe antennes werken, waarom de actieve antennes beter dan de passieve zijn, en verklaart de waargenomen verschillen in werking van diverse passieve antennes.

(8)

1

Introduction

1.1 Lighting

Lamps are so common that they are taken for granted. However, life without artificial lighting would look much different than what we are used to. Daily activities today do not depend on the weather conditions or the season we find ourselves in; this allows for the evolution of technology, science and art. There are millions of people staying at work after the sun has gone down in the winter months, performing their jobs and making our world possible. However, they rarely look up to examine the light emitters suspended above their heads.

About 20% of the world’s electricity consumption is used for lighting [1]. This criterium alone shows the major importance of good lamp design. Ad-ditionally, there are several other key features a good lamp should have, like good color rendering, which makes the colours of illuminated objects appear like they would under sunlight. The stability and the longevity of the light source are also sought-after properties. Having all this in mind, it becomes obvious that good lamp design is anything but simple - it often involves years of development and testing done by teams of experts in various fields.

Most of the light sources used today work on the principle of plasma tech-nology. Even though the LED technology is developing fast, discharge lamps are still most qualified for general lighting applications in large areas, for indoor and outdoor lighting [2]. Gas discharge lamps convert the electric power into light by means of an electrical discharge in the gas medium. A plasma is cre-ated in the gas volume between two electrodes [3] - this is a state of gas where

(15)

an appreciable amount of the atoms and molecules are excited and ionized. High intensity discharge (HID) lamps represent a lamp family that belongs to the larger group of discharge lamps [4]. These lamps operate at high pressure (over 1 bar) contained in a small volume and they usually contain mercury or sodium. High pressure surroundings ensure numerous collisions between elec-trons and heavy particles. A cumulative effect is the energy transfer from electrons to heavy particles, which results in high heavy particle temperatures, typically between 1400 and 8000 K. Consequently, the spectral power is dis-tributed over a large range of lines, while the resonance lines are typically self-absorbed.

Even though HID lamps operate at high pressures, the conditions during the cold ignition are quite different. Mercury pressure is substantially reduced and other metals undergo transitions back to their solid phase. Consequently, a medium is needed in a cold lamp, which one can convert into plasma. For this purpose noble gasses are commonly used.

1.2 Ignition

1.2.1 Discharge lamps

The ignition process in discharge lamps comprises of a sequence of events during which the starting gas in a lamp undergoes conversion into a conductive state and finally forms an arc [5]. The breakdown stage is the first one to take place; it encompasses the transition of the gas to a conductive state usually via the Townsend or the streamer mechanism. Subsequently the discharge needs to be stabilised. This stage entails the limiting of the current through the discharge, which would otherwise, due to the negative voltage-current characteristic, rise until the lamp would be destroyed. The various stages of discharge ignition have been portrayed in the case of low pressures in a homogenous electric field [6, 7] and are schematically shown in figure 1.1. The current-potential curve in the case of discharge lamps has been shown to have roughly the same complex shape [5].

In the beginning stage, denoted as I in the figure, when low-level voltage is applied to the electrode gap in a neutral gas atmosphere, primary electrons in the gas liberated by either photoemission from the cathode, radioactive decay or by cosmic radiation, move to the anode. This causes low-level current to 2

(16)

Introduction

Figure 1.1: The current-voltage characteristics of a DC discharge [5]. The current and voltage ranges for the are merely illustrative.

flow through the gap in this ’Geiger’ stage of ignition. No significant ionization takes place and the discharge relies on external sources of ionization. In order to increase the current, higher voltage must be applied. Consequently, the electrons are accelerated in the electric field, undergoing a series of elastic and inelastic collisions. New electrons are generated in inelastic collisions, and the resulting macroscopic current depends on their production rate.

When a high enough electric field is achieved, the discharge becomes self-sustaining and current breakdown is reached. This is shown as a crossing point between stage I and II in figure 1.1. At this point each electron in the electrode gap produces at least one secondary electron by means of ionization. Voltage at this level is called the breakdown voltage and it is a function of several parameters - gas type and pressure, electrode gap, electrode material and radius of curvature, voltage shape and slope.

Between the current and the voltage breakdown, in stage II, the current increases several orders of magnitude as a result of just a slight increase of voltage. This is partly due to the exponential form of the Townsend first ion-ization coefficient as a function of the electric field; this coefficient determines the ionization rate in the electrode gap. According to the Townsend theory, secondary electron emission processes take place as well, and are the primary cause of the steep current increase. This stage of the discharge development is called the ’Townsend discharge’ or the ’dark discharge’, as there is very little 3

(17)

light emitted.

At the point of voltage breakdown, crossing from stage II to stage III, the voltage across the electrode gap drops significantly and the discharge crosses to the next stage of its development - the ’subnormal glow discharge’ (stage III). The total ionization yield is an exponential function of the electric field, and the potential difference in the electrode gap is confined to the space near the cathode, which causes the electric field to grow in that region. As a conse-quence, the current through the discharge can increase even with a decrease of voltage across the gap. The total current is now determined by the discharge itself as well as the external electrical circuitry.

In the ’glow discharge’ stage (number IV in the figure), the voltage stays constant while the current through the discharge keeps increasing, due to the extension of the discharge over the electrode surface. Subsequently, during the ’abnormal glow discharge’ stage (stage V in the figure), the whole electrode surface is covered by the discharge and the increase of current can continue only by increasing the ionization rate. This can be done by increasing the potential difference, thereby effectively increasing the electric field in the electrode gap. When ample thermionic emission becomes possible as a consequence of cathode heating, another voltage drop with the increase of current is observed (stages VI and VII) and the transition to the ’arc discharge’, which is the final stage of the ignition process.

1.2.2 Breakdown

This thesis deals with the first stages (I and II) of the development of the dis-charge in a noble gas atmosphere, not only specific for HID lamps, but common in several other application areas. This is a transient phenomenon in which a body of neutral gas becomes conducting and it is referred to as breakdown. Simply put, breakdown entails the multiplication of electron avalanches. The multiplication factor M = γ exp(αd) is often used to characterize breakdown-associated phenomena. It gives the number of electron-ion pairs produced in the gap by the passage of one electron avalanche, according to the Townsend theory.

The product of gas pressure p and the length of the electrode gap d has been empirically identified as a parameter essential for expressing the break-down characteristics (breakbreak-down voltage) of an homogenous electrode gap by 4

(18)

Introduction F Paschen in the late 19th century [8]. The Paschen curves show the break-down voltage for a particular gas as a function of the pd product. At this voltage, the multiplication factor M is approximately equal to one. The obser-vations made by Paschen were explained by Townsend and the proposed break-down mechanism was named the Townsend mechanism. It should be noted, though, that the Paschen curve as well as the derivations in the subsequent section on Townsend mechanism are only valid for a parallel-plate (essentially 1-D) situation which is far from practical reality in discharge lamps and many other applications.

The increase in voltage causes the multiplication factor M to steeply rise, as the factor α - Townsend first ionization coefficient - is an exponential function of the electric field. M determines the formation time of the discharge. In the case when αd exceeds 20 [9], a localized space-charge region is created by the succession of electron avalanches in the electrode gap. This space charge focuses the electric field and drives the development of a conducting streamer channel [10].

More about the Townsend and the streamer mechanism can be found in the following two sections.

1.2.3 Townsend mechanism

Townsend was among the first scientists to study the variation of the current through gas discharges placed in a two-plane electrode system. He used UV irradiation to produce ample photoelectrons from the cathode surface to ensure a steady supply of initial free electrons at the beginning of the breakdown process [9]. He was the first to measure the current-voltage characteristic during the various stages of the breakdown process.

As the discharge crosses from the ’Geiger’ stage as described above to the stage of the ’Townsend discharge’, the steep increase in current is ascribed to the ionization of the gas by electron collisions. The increase in the number of electrons dn at the distance x from the cathode is given by [11]

dn = (α − η)ndx (1.1)

The parameters α and η are the ionization and loss coefficients and they typ-ically depend on the gas type, pressure and the electric field. α − η is the effective ionization coefficient. At the distance x = d, the number of electrons 5

(19)

is given by an exponential function

n = n0exp((α − η)d) (1.2)

Because of the exponential nature of electron multiplication process in the case when the effective ionization has a positive value, this process was named the electron avalanche. The corresponding current is given by

i = i0exp((α − η)d) (1.3)

In this case i0 is the primary photocurrent, while the n0 is the number of

primary electrons generated at the cathode.

During the ’Townsend discharge’ stage, the current was observed to rise faster than exponentially. This was explained by Townsend as a consequence of other processes, like inelastic collisions to form metastable or radiatively excited states of neutral atoms or molecules. Consequently, positive ions, photons and metastables were thought to be available in the gas to enable production of secondary electrons emitted either in gas or at the cathode. While scientists largely agreed with the theory of secondary electron emission from the cathode surface, most of them voiced their scepticism towards the same effect in the gas volume, arguing that the positive ions are not likely to gain enough energy to enable this process.

Including the secondary electron emission into the theory and neglecting η, the number of electrons reaching distance d per second is given by

n = n0

eαd

1 − γ(eαd_{− 1)} (1.4)

The parameter γ is the number of electrons released from the cathode per incident positive ion and it is called the second Townsend ionization coefficient. The current is therefore given as

i = i0

eαd

1 − γ(eαd_{− 1)} (1.5)

γ can represent one or more secondary electron emission mechanisms, as it has been shown that the resulting electron density has the same form when considering other secondary electron emission mechanisms.

At low overvoltages, the value of γ(eαd₋₁₎_{is close to zero and the}

aforemen-tioned expression takes the form of the equation derived for the case without 6

(20)

Introduction the secondary electron emission mechanisms. A singularity is reached when γ(eαd− 1) = 1. At this point, the current becomes indeterminate, and accord-ing to the original theory suggested by Townsend, this point signifies the onset of a spark. In general, eαd₁_{, and the condition can be written as}

γeαd= 1 (1.6)

The significance of Townsend sparking criterion has been widely discussed and the following conclusions have been drawn [9]. In the case when γeαd_{< 1,}

the current i is not self-maintained and ceases upon the removal of the pho-tocurrent i0. When the criterion is met, the multiplication factor M as defined

in the previous section is sufficiently large so that the total balance of ioniz-ing processes in one electron avalanche and subsequent cathode bombardment produce one secondary electron that will repeat the process. The discharge is then self-maintaining and can continue in the absence of the outside source of electrons. When γeαd_{> 1}_{, the ionization produced in successive avalanches is}

cumulative and the spark grows rapidly.

Ionization of the gas by electron collision and the secondary electron emis-sion from the cathode surface are chance phenomena, and the first and the second Townsend ionization coefficients fluctuate around the mean value. The product γeαd _{varies for individual avalanches. Consequently, there is an}

aver-age breakdown voltaver-age corresponding to the criterion γeαd_{= 1} _{and breakdown}

might be possible in some cases at lower voltages as well. However, due to the steep rate of change of the product γeαd _{with the voltage gradient, the}

breakdown voltage is generally very well defined.

From the breakdown voltage measurements, it is possible to relate the values of α/p and the E/p [9]. As α/p and γ were shown to be functions of E/p, it was not difficult to theoretically prove Paschen’s law previously empirically established. Given that the research was done in homogenous electric field, E = V /d, and the Townsend criterion can be rewritten in the following form

γ(V /pd)epdα(V /pd) = 1 (1.7)

It is clear that for a given product pd there is a particular value of V , and hence V = V (pd) [9].

(21)

1.2.4 Streamer mechanism

At atmospheric pressures and distances larger than a few centimeters, the Townsend mechanism ceases to be valid. The theory about breakdown pro-cesses at high pd values was developed by Loeb, Meek [12] and Raether [13]. They introduced the concept of space charge. The idea was that the stream-ers propagated (grew) in gas by ionizing the medium in front of them due to the high electric field present at their tips. The electric field, resulting from the charge separation at the streamer’s tip, was thought to be local and high enough to modify the background field imposed by the electrode system.

Where the Townsend theory fails

At pd values larger than 1000 Torr cm (around 1 Bar cm), the Townsend theory cannot explain the observed breakdown processes. In particular, there are three major inconsistencies between the observations and the theory devised by Townsend [9].

One, the formative time of the discharge (from onset to breakdown) is mea-sured in tens or hundreds of nanoseconds. The time it takes the ions to move to the cathode and create secondary electrons is much longer than the formative breakdown time.

Two, the influence of the Townsend secondary electron emission coefficient is not pronounced. This conclusion comes from the observation that the break-down voltage is independent of the type of cathode material.

Three, the discharges are not uniform and diffuse. They are thin and fila-mentary.

The transition from the avalanche mode to the streamer mode of dis-charge development is closely connected to the dis-charge present at the tip of the avalanche [14]. When the Townsend sparking condition is satisfied and the ion-ization is cumulatively produced in the successive avalanches, the space charge develops at the discharge tip. The space charge can grow sufficiently large to cause local amplification of electric field that in its magnitude surpasses the field imposed by the electrode system. The presence of this field can cause high ionization rates locally around the discharge tip and allow for new avalanches to start developing. The criterion for streamer onset is that the electric field produced by charge separation at the avalanche head must be approximately 8

(22)

Introduction equal to the electric field applied externally [9, 14]. From this condition the critical density of charge at the streamer tip can be deduced [9,13].

The total number of electrons at a distance x from the starting point of the discharge in non-uniform fields is calculated in a way that is a bit more general than the one presented for the uniform fields. It allows for the effective ionization coefficient to be a function of the position in the electrode gap.

n = n0exp

hZ x

0

(α − η)dxi (1.8)

α and η are given as a function of energy and gas type [15]. It has been found experimentally that the streamers can propagate in background fields lower than needed for inception, but that there still is a minimum background field needed for streamer growth (also called the stability field). It is of the order of MV/m for negative and a few hundred kV/m for positive streamers in atmospheric air.

Finally, the streamer onset and propagation criterion is just an extension of Townsend self-sustained avalanche criterion. First, the streamer onset con-dition must be satisfied, i.e. the avalanches have to produce space charge dense enough to cause local electric field greater than the background electric field. For this, the field in the vicinity of the starting electrode must first be fairly high [14]. Second, the background electric field must be larger than the minimum needed for sustaining streamer growth.

The streamer propagation mechanism

There are two basic types of streamers - positive (cathode-directed) and neg-ative (anode-directed) [7]. The basic mechanism of streamer propagation for both streamer types is ionization of the gas in front of the streamer tip due to the high electric field at the tip [7, 9, 14, 16–18]. This requires some minimum amount of free electrons to be present in the gas irrespective of the discharge formation. It is, however, common to find free electrons in the surrounding gas, originating from cosmic rays or radioactive materials commonly found in man-made structures.

The growth of a negative streamer does not require a pool of free electrons at its tip, as a negative streamer can supply its own free electrons from the region that has already been ionized. In the case of a positive streamer, the free 9

(23)

electrons drift towards the streamer body; consequently, the streamer cannot provide its own free electrons in front of its tip. There are other mechanisms needed to explain positive streamer growth.

A commonly included streamer propagation mechanism is photoionization of the gas molecules present locally around the streamer tip. This mechanism has been most often studied for discharges in air [7, 9, 14, 19–24]. The high-energy photons are thought to come from excited states at the streamer head and ionize neutral molecules in front of it. One should note that photoion-ization is possible only in gas mixtures, where the photon emitted from one excited neutral can ionize another. One should also note that the experiments in completely pure gasses are almost impossible to perform and that, at the same time, a very little amount of oxygen in argon atmosphere, for example, can appreciably change the discharge development [20,25].

These two mechanisms have in common that the ionization of the neutral gas takes place in front of the streamer head. Charge already contained in the streamer channel does not have to be transported to the front of the streamer (in the case of positive streamers) and collisionally excite the neutral atoms and molecules. The ionization is achieved by giving energy to free electrons already present in the electrode gap irrespective of the discharge existence or by photoionization. The expendability of the fast transport of charge explains the growth of streamers at velocities greater than the electron drift in the given conditions.

As the ionization happens in front of the streamer head, one could say that there is an active volume surrounding the tip in which the ionization processes are possible. This has been illustrated by Gallimberti in 1972 [11]. The volume covers the space where the electric field is high enough to sustain collisional ionization processes. Photoionization has a potential to expand the active volume, thus effectively speeding up the discharge growth. However, it has recently been shown that photoionization plays a lesser role than was previously believed [19,20].

1.3 Overview of the thesis

The goal for this project was to investigate several aspects of breakdown in noble gasses and HID geometry. The choice to do experiments in pure argon and xenon was made in attempt to decouple the various effects in the ignition 10

(24)

Introduction sequence in HID lamps, which is in itself a very complex process. Understand-ing breakdown in pure gasses is the steppUnderstand-ing stone for understandUnderstand-ing more complex processes involving mercury and metal vapour, but also the physics behind breakdown in gas mixtures such as air.

We set out to examine a few aspects of pulsed breakdown in a pin-pin ge-ometry, such as the influence of dielectric surfaces and the role that metastable atomic argon plays in ionization processes during pulsed breakdown in the gas volume. Both pulsed surface and gas discharges have already been researched, but there is knowledge lacking for a full description, which is especially true for surface discharges.

Experimental results of the surface discharge experiment are presented in Part I of the thesis. A pin-pin geometry was used, with a flat dielectric surface placed close to the electrode system in a way where the electric field direction was parallel to the dielectric surface we subsequently introduced. The elec-trodes were not in contact with the dielectric. Both discharges in gas volume and on the surface were obtained and their velocity measured and compared. The comparison of surface and gas discharges presented here is of value for both the lighting applications and for the pool of general knowledge on surface discharges, as the comprehensive theory on this topic has still not been devel-oped. In lighting applications, this comparison aids in better understanding of non-aided and antenna-aided lamp ignition. The work done on antennas is presented in Part IV of the thesis.

Part II brings results of simulations of pulsed breakdown in argon in a pin-pin geometry. A useful property of research done using simulations is that one is often able to separate various effects present in a research topic that are practically inseparable in experiments. This is what we have done in this case. In order to probe the influence and the role of argon metastable atoms during pulsed breakdown, simulations were performed that compare breakdown properties when stepwise ionization is and is not included in the calculations.

Previously rather poorly researched topic of high-frequency AC breakdown in the frequency range where electron losses are drift dominated is presented in the three chapters of Part III. It has been previously empirically proven that substituting pulsed voltage with the AC signal significantly reduces the ignition voltage in HID lamps [26]. This part of the thesis brings results of experiments and simulations designed to discover the main characteristics of AC-driven breakdown, identify the important processes and deduce their respective roles.

(25)

Part IV of the thesis is the most application-driven part of the text. Effects of metallic structures on the outside of the lamp burner were examined for their effects on the breakdown process. This part brings results of the experiments, compares them to the results of non-aided breakdown processes in HID lamps presented in Part III and offers a theory to explain the differences between the breakdown observed using guided discharges with different antenna potentials.

The thesis is completed with concluding remarks and a general outlook.

(26)

Part I

Pulsed surface and volume discharges

in argon

(27)

(28)

2

Speed of streamers in argon over a flat surface of a

dielectric

Abstract. A pin-pin electrode geometry was used to study the velocities of streamers propagating over a flat dielectric surface and in gas close to the dielectric. The experiments were done in an argon atmosphere, at pressures from 0.1 to 1 bar, with repetitive voltage pulses. The dielectric surface played a noticeable role in discharge ignition and propagation. The average speed of the discharge decreased with higher pressure and lower voltage pulse rise rate. Moreover, it was higher when the conductive channel between the electrodes was formed over the dielectric rather than through the gas. Space resolved measurements revealed an increase in velocity of the discharge as it travelled towards the grounded electrode.

This chapter was published in Journal of Physics D: Applied Physics J. Phys. D: Appl. Phys. 42 (2009) 015211 in a slightly altered form.

(29)

Part I.

2.1 Introduction

High pressure discharges (0.1 to 1 bar) that travel along insulator surfaces in shapes of thin, ionized channels (also known as streamers) already have the attention of scientists. Either they want to avoid them, as is the case in high power electronics [27], or they want to use them in removal of unwanted electric charges from the surfaces of aircrafts in flight, photocopying, handling waste, UV generation [28, 29], flow control [30] or water treatment [31]. The lighting industry also has an interest in understanding the role of streamer discharges in lamp ignition in the presence of dielectric wall materials.

Most of the gathered knowledge comes from electrical engineers, who have been troubled by this phenomenon for years [32–49] because material interac-tion with the streamer development is of significance when it comes to insu-lation performance. For the most part, their research was done in air and in atmospheric pressure.

2.1.1 Breakdown in gas and on insulating surface

The inception conditions for surface and gas discharges are quite similar -the necessary requirements are a free electron, an external electric field whose strength is above the critical value, and a sufficient distance between the elec-trodes [9]. The initial electrons may be created by natural ionization or be left over from previous discharges. Following the inception stage, once a space charge cloud has developed, the local electric field becomes strong enough to start new avalanches. The local electric field and the photoionization are the mechanisms suggested responsible for discharge growth [7]. Adding an insu-lator surface to the process makes it far more complex. It has been observed that the dielectric strength of the gaps change significantly [44,47,50], presum-ably because of the changes of ionization coefficients and transport parameters induced by the presence of the dielectric surface. Even though the cornerstone of understanding of surface discharges has been laid in the field of breakdown in gasses, we still do not have a complete apprehension of the physics of this phenomenon.

Most authors agree that the presence of the insulating surface near the growing discharge causes the distortion of the electric field [28, 34–39, 41–47, 50–52], which influences the electric field at the streamer tip and causes the discharge to attach to the insulator surface. Also, the effective ionization and 16

(30)

Surface and volume discharge velocities attachment rates seem to be affected in different ways during the discharge growth, resulting in changes in growth characteristics.

2.1.2 Theory of desorption flashover

In 1977, a theory to explain the dramatic decrease of dielectric strength of gas gaps in the presence of dielectric surfaces was laid out by Avdienko and Malev [50]. At that time, the competing hypotheses were surface charge ac-cumulation and the discharge formation in the adsorbed charge layer on the dielectric surface. They divided the flashover phenomena that had been ob-served until that time with respect to the dielectric conductivity. The thermal flashover was observed at surfaces with higher conductivity, where the dielec-tric strength of the gap decreased with the increase of temperature, irrespective of the surface shape. The thermal flashover theory makes use of the heat con-duction properties of the dielectric surface.

The dielectrics with lower electrical conductivity did not show the same be-haviour with the increase in temperature, but the dielectric strength of the gap did depend on the shape of the insulating surface. The theory [50] stated that the flashover on this kind of surfaces (the desorption flashover) was connected to gas desorption from the dielectric surface, or outgassing, when subject to electrical stress. By 1977, this effect has already been observed (see references in [50]), but also shown, measured or deduced from calculations later [53–55]. The work described in [54] showed that the desorbed gas is usually not the background gas in the experiment, but most probably gas adsorbed during the handling process (nitrogen, water vapour etc.). Both [54, 55] agree on the importance of outgassing in surface breakdown process.

The phenomenon of desorption flashover was thought to consist of three steps [50, 53, 56, 57]. First, high stress is applied to the insulator under which the insulating surface acquires positive charge. A hopping electron model was used to describe this phenomenon, where secondary electrons are produced by electron impact at the insulating surface. The initial electrons were assumed to be present due to natural ionization. In [57], the authors suggest that this mechanism is self-limiting and that the current of the avalanche is controlled entirely by an emission site which emits initial pre-flashover electrons. The suggested site was the triple point, the junction of the cathode, dielectric surface and the vacuum, also proposed previously [56]. Secondary electron emission by electron impact was also investigated by [43].

(31)

Part I.

Next, the outgassing of the dielectric surface occurs. The electrons moving almost parallel to the dielectric surface cause gas molecules to be desorbed from the surface with great efficiency, between 100 and 200 molecules per incident electron [50]. The reason for this great efficiency is the electron trajectory directed along the insulating surface, not perpendicular to it. This value was under dispute, and various values can be found in the literature (for example, 4-8 molecules per electron in [53]). The next stage of discharge development is free electron generation in the desorbed gas layer by ordinary volume processes, subsequent charge multiplication, and breakdown (flashover).

The physical properties of the dielectric and vacuum conditions were thought not to have practically any influence on the characteristics of desorption flash-over, since to form a discharge, one needs only to desorb less than one mono-layer of adatoms. It is very important to notice that the theory proposed in [50] relied on the fact that the electric field in the system was parallel to the dielec-tric surface. In the case of guided discharges, i.e. the discharges where there is an electrode present parallel to the insulating surface [51], this condition is not met. In this case, the surface charge was expected to take a notable part in the breakdown process [50].

2.1.3 Other observations

In 1982, a simple theory was developed to explain the behaviour of the break-down voltage of a gas gap situated at the axis of a dielectric cylinder, when the radius of the cylinder is varied [58]. The theory was postulated on the premise that breakdown will occur when the product of electron density and the average drift velocity reaches a certain value. The results of the predicted breakdown voltage were in good agreement with the experimental data.

The theories nowadays are differently focused. Several authors have shown [35, 44] that the streamers usually propagate in two modes - there is a fast component over the dielectric surface and a slow one in ambient air. The most obvious explanation for this phenomenon is the modification of the electric field due to the dielectric permittivity. Other explanations were offered as well, for example that the ionization rate is altered in presence of an insulator because of electrons emitted from the dielectric surface by photoemission [35–37,41,43,46] and detachment of negative ions [38, 40]. The authors suggested that these processes speed up the propagating streamers over a dielectric surface, even though they have to compete with losses due to electron and ion attachment to 18

(32)

Surface and volume discharge velocities the surface. It has been shown that the dynamics observed in surface discharges cannot be attributed only to the boost of bulk processes due to electric field modification near the dielectric surface [59–61].

In order to unravel the influence of an insulating surface on the basic prop-erties of streamers, extensive measurements were done, determining streamer velocities that depend on various factors [37–39, 46, 62]. The velocities varied greatly with the applied voltage and the electrode configurations.

Charge deposition during the formative avalanches of streamer corona was investigated, modelled [34] and measured [38, 42, 45, 63–65], along with the streamer velocities [41]. The measurements were done for various shapes of the dielectric test objects as well as for different electrode geometries, the most popular being the pin-plate geometry. The charge deposited by corona avalanches on the dielectric surface modified the growth process and lead to a reduction of the total charge generated in the corona. Allen and Faircloth [41] saw that the net charge deposited over the insulator surface was small compared to the injected corona charge. Surface charge density distribution followed the profile that the streamers form, and it increased the probability of the consecutive breakdown. The densities close to the electrodes were high, which could contribute to breakdown probabilities via high local electric fields and associated supply of initiating electrons. There was a negligible increase in deposited charge after successive coronas.

The deposited charge was also shown to hinder inception [34] or not influ-ence it at all [40], but promote discharge growth [40]. The effect of traps in the dielectric surface was modelled [66] and found to have a large influence on the flashover properties, as it affects the charge transport. By comparing measurements done with and without UV irradiation, it was proven that in the case of nitrogen discharges, surface charge influences the discharge growth [67].

2.1.4 Guided discharges

According to Fouracre et al [51], guided discharges are discharges on the dielec-tric surface when there is an electrode (usually grounded) present on the other side of the dielectric. Many authors have done experimental [28, 51, 68–72] or theoretical research [29,73] into the subject.

The electric field configuration and strength in this case is much different than in the case we have been examining until now - where the electric field lines 19

(33)

Part I.

are directed parallel to the dielectric surface. There are numerous differences between these two arrangements, but they all come down to changes in the relative influences of the effects present in both electric field configurations.

First, the dielectric permittivity plays a bigger role in the guided discharges, and its change is expected to have a greater effect on the discharge properties, especially during the growth stage. The expected effect of the permittivity was shown by [28, 51]. As a result, the bulk gas ionization processes are ex-pected to play a greater role in the guided surface discharges than in the case when the electric field is parallel to the insulating surface. Consequently, the other surface processes become less important in the overall picture. This was demonstrated in simulations done by [29]. The bulk processes played by far the most important role in the discharge growth, contributing to the overall ionization process over one order of magnitude more than the next process in line - the ionization by sheath-accelerated electrons. This effect was followed by secondary electron emission from the insulating surface and subsequently photoionization. The difference in contribution between bulk ionization and photoionization was more than three orders of magnitude.

Second, the desorption theory described previously cannot explain the phe-nomenon well, as the relative importance of surface charges grows with the an-gle between the electric field and the insulating surface [50]. The importance of surface charges in guided discharges was demonstrated by [51,68,70,74]. They concluded that the velocity of the surface discharge is strongly dependent on the charge density previously deposited on the surface. The authors of [68] even demonstrated a self-propagating surface discharge on sufficiently pre-charged surfaces. The velocities of surface discharges were found to be greatest if the surface charge of opposite polarity was pre-deposited on the surface [51,68,70].

2.1.5 Lighting industry

The lighting industry is currently taking an interest in surface discharges as well, as these appear to play a significant role in the ignition processes of plasma-driven lamps. This was shown in several recent investigations [26, 75, 76], but never systematically studied. Knowledge about surface effects in lamp burners could be used for modifying lamp ignition voltage, which is an important parameter in lamp starting: ballasts are designed to provide sufficient voltage to ignite a lamp. However, because of the stochastic nature associated with breakdown [77], a lamp will not ignite at the same voltage 20

(34)

Surface and volume discharge velocities every time. Studies done on this subject showed large variations in ignition voltage [75,77]. This makes lamp ignition uncertain unless its ballast provides voltage much higher than actually required for ignition. Lowering the ignition voltage or reducing its large spread would open possibilities for improvement of lamp characteristics.

The lighting industry and the electrical engineering community share an interest in surface discharges. A fundamental difference between them is that electrical engineers limit their research to atmospheric pressure air conditions, while the lighting industry is much more interested in discharges in noble gasses, in a variety of pressures. The existing knowledge is mainly on at-mospheric air discharges, and there are essential differences between air and argon kinetics (e.g. detachment of electrons from negative ions). This is why we cannot directly use any theory for air in case of argon. Even so, in our experiments with argon, we have observed the same phenomena as previously reported for air (e.g. propagation in two modes - a fast component over the dielectric surface and a slow one in ambient gas) and reviewed the proposed physical processes that could be responsible for observed effects with respect to gas type. In addition, in this chapter we present a small selection of modelling results that were obtained for discharges in argon.

Appearance of the discharge in argon near a flat dielectric has already been reported on [52]. In this chapter we present new data on pulsed discharge velocities over a dielectric surface in argon atmosphere, obtained with a high speed iCCD camera and two different dielectric materials. We show the results of average speed measurements, as well as space resolved velocities.

2.2 Experiment

The setup is shown in figure 2.1. The negative output of a DC high voltage supply was used to charge a 1 nF capacitor through a 25 MΩ resistor, and as the fast switch closed, a positive pulse was brought to the top electrode, while the bottom electrode was grounded. The fast semiconductor switch was necessary to ensure short rise times of our voltage pulse and low jitter. Jitter is one of the biggest challenges encountered in the study of pulsed breakdown, streamers or corona discharges [78] because it keeps from obtaining sufficient time resolution in the experiment. We used a low-jitter scheme, based on the research of Briels and van Veldhuizen [78–80]. The setup shown in figure 21

(35)

Part I.

Figure 2.1: Schematic view of the electrode and dielectric geometry and the pulsed power circuit. We used an iCCD camera to photograph the discharge that formed between the electrodes. The iCCD camera was positioned perpendicular to the side of the dielectric, so that the electrode-dielectric system appeared in the photographs in the same way it is shown on the schematics.

2.1, which we used in our experiments, is a modification of their scheme. We used the Behlke HTS-361 semiconductor switch with trigger jitter of 0.1 ns, maximum withstand voltage of 30 kV and current limit of 60 A. Resistors R6

limited the current through the switch and were responsible for pulse rise time. Total value of 1kΩ allowed for a fixed rise time of 150 ns.

Since the total rise time was kept constant, the rise rate had to be changed by adjustment of the peak voltage of the pulse, which means that for example for a 100 V/ns rise rate, we had to use a pulse with peak voltage of 15 kV. 100 V/ns was the fastest pulse we used. Experiments were also done with 67 V/ns and 47 V/ns pulses, which translates to 10 kV and 7 kV peak voltage, respectively.

Rod shaped tungsten electrodes were used, 0.65 mm in diameter. A flat piece of dielectric material (130 × 40 × 8 mm) was placed in the vicinity of the electrode system, as shown in figure 2.1. We used two dielectric materials -BK7 optical glass (rbetween 2.3 and 4.5) and aluminum oxide (r= 9.1). The

surface of aluminum oxide was slightly more rough than that of the BK7 piece, but based on the work of Tan et al [46], we expected the surface roughness not 22

(36)

Surface and volume discharge velocities to have an influence on the experiment.

The distances in the system were not fixed, which allowed us to choose spacing between the tips of the two electrodes and between the electrodes and the surface of the dielectric. Since we wanted to compare the velocity of the discharge over the dielectric surface and in the gas volume between the electrodes, we chose to present three geometries that provided the conditions for acquiring both, sometimes even at the same time. The parameters are given in table 2.1.

We investigated the pressure dependence of measured velocities, that is why the electrode system was enclosed in a vacuum vessel. A Pfeiffer Vacuum rotary vane DUO 10 pump was used to control argon pressure in the vessel, in range from 100 mbar to 1 bar.

One trigger unit was used to trigger the camera and the electrical system. We examined the influence of trigger frequency on the discharge in the range from 0.2 to 25 Hz. Our investigations have shown similar behaviour for all frequencies, with breakdown voltage falling as we increased the frequency [81]. Here we present data for 1 and 25 Hz for detailed velocity measurements.

Discharge velocities were measured by means of fast photography. We used an intensified CCD (iCCD) 4Quick Edig camera from Stanford Computer Op-tics Inc., aligned to image the electrodes and the side of the dielectric. Pho-tographs were taken using different delay times with respect to the trigger pulse, to image different stages of the developing discharge. The exposure time was always 5 ns. It should be mentioned that we could never take multi-ple pictures of the same discharge. Even so, with jitter of nanosecond order of magnitude, we were able to clearly distinguish between the photographs taken at different delay times.

2.3 Appearance of the discharge

Figure 2.2 shows photographs of discharges for three different geometries used in our experiments. All three pictures were taken at 400 mbar and pulse rise rate 100 V/ns, for the same dielectric material (BK7). One notices streamers, which is expected for a fast pulse discharge at this pressure.

The discharge propagated through the gas volume between the electrodes and reached the dielectric surface in all three cases. The mechanism responsible 23

(37)

Part I.

Figure 2.2: Photographs of discharges for three different geometries, specified in the right bottom corners. 20/12 denotes the geometry with 20 mm spacing between the electrodes and 12 mm between the electrodes and the dielectric, 20/7 marks the 20 mm - 7 mm and 30/7 specifies the 30 mm - 7 mm combination. The dielectric and the electrodes are shown in the figure, with top electrode as the anode, and the bottom one grounded. Pictures were taken with exposure times of 5 ns, at the moment very close to completion of the conductive channel between the two electrodes, for 400 mbar and 100 V/ns rise rate. The photographs are presented in false colour.

for the discharge choosing to propagate towards the dielectric is connected to the shape of the initial electric field, before the start of rapid streamer growth. It is well known that the dielectric material reacts to the applied electric field by aligning its dipoles, thereby lowering the field in its volume. As a result, the electric field outside the dielectric is increased. The difference between the electric field when the dielectric plate is close to the electrodes as opposed to the geometry in which the dielectric is further away has already been demonstrated [52]. Potential distribution was obtained from a model and it clearly showed a steep potential gradient between the electrode tip and the dielectric, which can explain the inclination of the discharge to move towards the dielectric.

One can see that the propagation path of the discharge changed as we moved the dielectric or changed the spacing between the electrodes. Even though there were streamers present in the gas volume as well as on the dielectric surface for all three geometries, the difference between these three discharges is in the location of an ionized channel that ultimately connected the two electrodes. 24

(38)

Surface and volume discharge velocities

Table 2.1: Geometry parameters. s denotes the spacing between the tips of the electrodes, and d the distance between the electrodes and the dielectric. Even though a series of combinations of s and d values was examined, we show discharge velocities for only three combinations of these parameters, as they represent three different modes of discharge propagation.

cases s [mm] d [mm] the discharge propagates . . .

a 20 12 in gas

b 20 7 in gas and over the dielectric surface c 30 7 over the dielectric surface

For a dielectric placed 12 mm from the electrodes (the left most picture in figure 2.2), the ionized path was formed between the electrodes, through the gas volume. On the other hand, when the dielectric was placed closer, at 7 mm distance from the electrodes, and they were placed 30 mm apart (the right most picture in figure 2.2), the ionized channel formed over the dielectric surface.

There was a mid point in the configuration, or mid region, in which the discharge propagated both over the dielectric surface and through the gas vol-ume, and formed two channels roughly at the same time (the middle picture in figure 2.2). This was again seen in a geometry with electrodes and dielectric moved about 50% further apart. When the dielectric was moved further away at same electrode spacing, the discharge made a channel only through the gas. The same behavior was observed in the whole pressure range.

2.4 Average velocities

The average velocity of the developing discharge was determined from the ratio of the time it takes a discharge to cross the gap between the electrodes and the length of the path it travels. We found the starting time in the following manner: first, we determined the time delay from the trigger signal after which there was 20% chance out of 20 discharges to observe light emission from the discharge. Let us call this time a. After that, we determined time b, after which there was 80% chance to make such an observation. The starting time was calculated as an arithmetical average of times a and b. We did the same for the moment when the ionized channel was completed - we found the delay times where there was 20% and 80% chance to observe a complete channel between the electrodes and calculated their arithmetical average. The length 25

(39)

Part I.

of the path crossed by the discharge was measured from the photographs of the discharges.

2.4.1 Measuring conditions

Velocities were measured at ten pressures from 0.1 to 1 bar, for three voltage rise rates: 47, 67 and 100 V/ns, and two dielectric materials. Photographs of successive discharges were taken with 5 ns gate time at two trigger frequencies: 1 Hz and 25 Hz. We discovered that the 25 Hz discharges suffered from con-siderably less jitter, thus making the measurements much more reliable and easier to perform. Error bars at 1 Hz were roughly 5 to 25 times larger than for 25 Hz.

Furthermore, velocities at 1 Hz were 55% to 65% larger than their 25 Hz counterparts at 0.1 bar, with this difference approaching zero as we increase the pressure. This effect can be explained as follows: It has already been shown [38,42,45] that charge remains on the dielectric surface for a long time after a discharge, even for days in cases of very good insulators. The results show that after minutes of waiting time, the leftover charge preserves the pro-file of the streamers that were involved in its deposition. This means that throughout our experiment, which ever trigger frequency we used, we had left-over charge remaining in the gas volume and on the dielectric surface after every discharge. As the density of the leftover charge will drop with time, op-erating the experiment at 25 Hz will cause more leftover charge to be present in the gas volume and on the dielectric surface than at some lower trigger fre-quency. This leftover charge plays a twofold role. First, it serves as a source of free electrons necessary for the initiation of the discharge. More electrons causes smaller statistical lag time, which makes the discharge less prone to jitter and easier to predict. Second, lots of leftover electrons can serve as a shield, screening the applied potential at the charged electrode. Smaller E/N means lower discharge velocities [82]. Both effects were seen in our experiment. The effect of leftover charge was also pondered upon previously [34,40,41,50]. All measurements were done for 1 Hz as well as for 25 Hz, and all of them showed the same behavior. For the sake of clarity and reliability, we present only the results of the 25 Hz measurements in the rest of this chapter.

(40)

Surface and volume discharge velocities

Figure 2.3: Relationship between average discharge speed and electrode-dielectric geometry. The geometries are noted as 20mm/7mm and 30mm/7mm, which stands for 20 or 30 mm spacing between the electrodes and 7 mm distance from electrodes to the dielectric. The experiments were done on BK7, at 25 Hz, 67 V/ns.

2.4.2 Effect of pressure and geometry

We measured the difference in velocities when the discharge travelled through the gas or over the dielectric. Figure 2.3 shows velocities for BK7, constant voltage rise rate of 67 V/ns (10 kV maximum voltage) and two geometries mentioned in table 2.1.

First, it can be seen that the velocities decreased with rising pressure. As pressure rises, the mean free path of electrons is decreased, effectively slowing down the discharge development.

Second, one can observe higher velocities when the discharge formed an ionized channel over the dielectric surface as opposed to when the channel was formed in the gas. This difference can be seen clearly for the geometry (20mm/7mm) that features both ways of the channel’s formation; in this case the two velocities were measured simultaneously. The question is why the propagation over the dielectric surface is faster than through the gas. Many scientists assumed that this is the result of the dielectric properties of the material [28,34–39,41–47,50,52]. A combination of permittivity of the dielectric and accumulated negative charge on its surface was also suggested [40,41] to be responsible for an increase of the ionization rate on the dielectric surface, that 27

(41)

Part I.

Figure 2.4: Average discharge velocity as a function of dielectric material and voltage pulse rise rate. 25 Hz, 30mm/7mm geometry. The discharge formed a conductive channel over the dielectric surface.

speeds up the discharge. Finally, there is the desorption theory [50, 53–55, 57] that predicts a joint influence of secondary electron emission from the surface, outgassing and subsequent breakdown in the evaporated gas just above the insulating surface.

2.4.3 Influence of dielectric material and voltage rise rate

In figure 2.4, we compare average discharge velocities over the dielectric surface for two dielectric materials. The velocities along their surfaces were about the same, even though their relative dielectric constants differ by roughly a factor three. The material with lower r should produce smaller modifications of the

electric field in its vicinity, and therefore have a smaller influence on the speed of the discharge. Small differences (with maximum of 40% at 100 mbar) were observed in lower pressures, but the material with the lower dielectric constant (BK7) exhibited higher velocities.

The local electric field is also affected by the appearance of streamers. To find out in which way, we adapted a model, described in detail in the next chapter. It is a 2D fluid model, suitable for simulating nonuniform electric fields. Five species were included in the calculations - argon atom, argon ion, molecular ion, electron and argon metastable at 11.6 eV. We used the model 28

(42)

Surface and volume discharge velocities to simulate the starting phase of a discharge in argon at 700 mbar, beginning from the gas phase with low initial concentration of charged species, up to the formation of an ionized connection between the electrodes. The fact that this is not a 3D model highly specialized for our experiment gives rise to errors in absolute values of its results. The advantage is that this model can be used to reproduce the whole process from the low initial particle densities, which is typically the case in gasses, until the moment the ionized channel forms between the electrodes. The fact that the model is 2D causes the electric fields at sharp electrode tips and at the streamer heads to be lower than in reality, but it gives correct trends and correctly describes the behaviour of the discharge.

The simulations showed that the tip of the streamer was nearly at the same potential as its origin, the charged electrode. Maximum of 20% drop in potential was observed across the streamer length, just prior to the formation of the connection between the electrodes. This can easily be explained by the fact that a streamer is essentially an ionized, conductive channel. As streamer heads in argon are very narrow (a 5 kV streamer at atmospheric pressure is a fraction of a millimeter wide [78]), a high electric field is present in the area around streamer tips, with values that exceed the electric field imposed by electrode potential. Compared to the effects of the space charge, that completely alters the local profile of the electric field, the modifications caused by the dielectric property of the insulating material will have a small impact on streamer velocity along its surface.

Impact ionization of electrode surface by heavy particles in the discharge is one of the processes often considered relevant for fast propagation of discharges. However, it is by orders of magnitude too slow, as the whole ignition process in our experiment takes up to 150 ns.

In this circumstances, one has to consider other sources of electrons that can enable fast streamer propagation. As has been already suggested, leftover charge can play a role in streamer growth [40, 41]. The desorption theory also predicts extra free electrons near the insulating surface [50, 53–55, 57]. In any case, it has been shown that in non-guided discharges in rare gasses the expansion speed is governed by both drift electrons and free electrons generated by a short-range source in a narrow layer around the channel surface [59].

On the other hand, photoemission of electrons from the dielectric surface can be viewed as a mechanism partly responsible for the fast propagation.

Breakdown processes in HID lamps : exploration of various key aspects

Breakdown processes in HID lamps : exploration of various

key aspects

Breakdown processes

in HID lamps

Exploration of various key aspects

Summary

Samenvatting

Contents

1

Introduction

1.1

Lighting

1.2

Ignition

1.3

Overview of the thesis

Part I

Pulsed surface and volume discharges

in argon

2

Speed of streamers in argon over a flat surface of a

dielectric

2.1

Introduction

2.2

Experiment

2.3

Appearance of the discharge

2.4

Average velocities